Instructions to use moonshotai/Kimi-K3 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use moonshotai/Kimi-K3 with Transformers:

# Use a pipeline as a high-level helper
from transformers import pipeline

pipe = pipeline("image-text-to-text", model="moonshotai/Kimi-K3", trust_remote_code=True)
messages = [
    {
        "role": "user",
        "content": [
            {"type": "image", "url": "https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/p-blog/candy.JPG"},
            {"type": "text", "text": "What animal is on the candy?"}
        ]
    },
]
pipe(text=messages)

# Load model directly
from transformers import AutoModel
model = AutoModel.from_pretrained("moonshotai/Kimi-K3", trust_remote_code=True, device_map="auto")

Inference
HuggingChat
Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use moonshotai/Kimi-K3 with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "moonshotai/Kimi-K3"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "moonshotai/Kimi-K3",
		"messages": [
			{
				"role": "user",
				"content": [
					{
						"type": "text",
						"text": "Describe this image in one sentence."
					},
					{
						"type": "image_url",
						"image_url": {
							"url": "https://cdn.britannica.com/61/93061-050-99147DCE/Statue-of-Liberty-Island-New-York-Bay.jpg"
						}
					}
				]
			}
		]
	}'

Use Docker

docker model run hf.co/moonshotai/Kimi-K3

SGLang

How to use moonshotai/Kimi-K3 with SGLang:

Install from pip and serve model

# Install SGLang from pip:
pip install sglang
# Start the SGLang server:
python3 -m sglang.launch_server \
    --model-path "moonshotai/Kimi-K3" \
    --host 0.0.0.0 \
    --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "moonshotai/Kimi-K3",
		"messages": [
			{
				"role": "user",
				"content": [
					{
						"type": "text",
						"text": "Describe this image in one sentence."
					},
					{
						"type": "image_url",
						"image_url": {
							"url": "https://cdn.britannica.com/61/93061-050-99147DCE/Statue-of-Liberty-Island-New-York-Bay.jpg"
						}
					}
				]
			}
		]
	}'

Use Docker images

docker run --gpus all \
    --shm-size 32g \
    -p 30000:30000 \
    -v ~/.cache/huggingface:/root/.cache/huggingface \
    --env "HF_TOKEN=<secret>" \
    --ipc=host \
    lmsysorg/sglang:latest \
    python3 -m sglang.launch_server \
        --model-path "moonshotai/Kimi-K3" \
        --host 0.0.0.0 \
        --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "moonshotai/Kimi-K3",
		"messages": [
			{
				"role": "user",
				"content": [
					{
						"type": "text",
						"text": "Describe this image in one sentence."
					},
					{
						"type": "image_url",
						"image_url": {
							"url": "https://cdn.britannica.com/61/93061-050-99147DCE/Statue-of-Liberty-Island-New-York-Bay.jpg"
						}
					}
				]
			}
		]
	}'

Docker Model Runner
How to use moonshotai/Kimi-K3 with Docker Model Runner:
```
docker model run hf.co/moonshotai/Kimi-K3
```

Kimi K3 (2.78T) running on a 128 GB MacBook Pro at ~1 tok/s, experts streamed from four SSDs

#185

by konstantinnikol - opened about 7 hours ago

Discussion

konstantinnikol

about 7 hours ago

Not a quant release, just a data point for anyone wondering whether K3 runs on a laptop at all.

Machine: MacBook Pro M5 Max, 128 GB. The trunk stays in memory and the routed experts (1.44 TB at 4-bit) are streamed from NVMe: the internal SSD plus up to three Thunderbolt 5 enclosures, each holding a byte-identical replica, with every expert read split across the drives. Output is identical at every rung.

one drive: 0.55 tok/s
+1 enclosure: 0.75
+2: 0.89
+3: 0.96 (1.75×), and 1.00 steady over 512 generated tokens, 1.13 over 128

Engine (a llama.cpp-derived fork), manifests and every run: https://github.com/argonautlabsai/deltafin (k3-public-bench). The storage layout and instruments are in argonautlabsai/argodrive.

CoderX1997

about 5 hours ago

why would u do this to ur self... 1 token / s is torture...

konstantinnikol

about 5 hours ago

why would u do this to ur self... 1 token / s is torture...

Fair, it's not a chat speed. The point was the existence proof and the curve: a 2.78T model on a laptop, same output as a resident run, and every SSD you add buys a measurable step.

In practice it runs a batch of document jobs overnight.

CoderX1997

about 1 hour ago

assuming it gets 2x faster per SSD, it takes u need another 6 drives to get a useable speed... not to mention, the processor can still bottle neck...

Upload images, audio, and videos by dragging in the text input, pasting, or clicking here.

Tap or paste here to upload images

· Sign up or log in to comment